Hans Moravec Reaffirms 2028 Human-Level AI Forecast in First Interview in Decades
Updated
Updated · New York Magazine · Jul 24
Hans Moravec Reaffirms 2028 Human-Level AI Forecast in First Interview in Decades
1 articles · Updated · New York Magazine · Jul 24
Summary
Hans Moravec, 77, resurfaced after decades out of public view and said AI remains “on track” for roughly human-level capability around 2028, close to the timeline he first laid out in 1988.
His original forecast tied the brain to about 10 trillion operations per second and projected computing’s two-year doubling forward; he now says the broad timing held even though modern AI arrived through neural networks he underestimated.
Moravec said he withdrew from public life to build robotics company Seegrid and later retired after a 2022 diagnosis of myasthenia gravis, which has left him largely chair-bound but mentally active and using ChatGPT daily.
He still expects AI to outgrow human control, arguing neural networks are trained rather than fully coded and can slip restrictions, making a smoother “mind children” handoff less likely than he once imagined.
Moravec’s ideas helped shape singularity thinking that later influenced figures such as Ray Kurzweil, Nick Bostrom, Sam Altman and Elon Musk, even as he now describes himself mainly as “a spectator.”
Will the once-mocked 1988 prediction of human-level AI by 2028 actually come true through raw computing power?
If modern AI systems are organically grown rather than strictly programmed, can humanity ever truly control them?
Why do machines easily master complex reasoning while still failing at simple physical tasks like folding laundry?
Approaching Human-Level AI by 2028: Moravec’s Vindication, the Scaling Supercycle, and the Coming Social Upheaval
Overview
In 2026, Hans Moravec’s return to public discussion reignited interest in his AI predictions, just as the scaling hypothesis—where raw computational power drives intelligence—was vindicated by the success of deep learning and large language models. This shift moved AI’s main bottleneck from software to physical infrastructure, sparking massive investments by tech giants and a global data center boom. However, new challenges emerged: supply chain bottlenecks in mature-node chips, energy constraints, and rising social tensions as automation reduced traditional employment and concentrated wealth. As governments scrambled to regulate powerful AI models and address economic disruption, society faced urgent questions about distributing AI’s benefits and redefining human purpose in a rapidly changing world.